Reducing borders of k-disjunction free representations of frequent patterns

  • Authors:
  • Marzena Kryszkiewicz

  • Affiliations:
  • Warsaw University of Technology, Warsaw, Poland

  • Venue:
  • Proceedings of the 2004 ACM symposium on Applied computing
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

A number of concise lossless representations of frequent patterns were proposed. Except for closed patterns, all other representations of patterns consist of a main component and a border. Recently, a unifying framework was introduced that treats border representations as particular cases of so called k-disjunction free representations. It was shown that careful splitting of borders into subgroups allows deletion of some of such subgroups without making the representations lossy. In this paper, we propose a new method of border reduction. Our method consists in identifying patterns in the main representation's component that uniquely determine a possibly maximal subset of patterns from the group under reduction. Such itemsets are redundant and can be deleted from border's groups. The performed experiments show that the new method reduces border representations by up to two orders of magnitude.